What's next after CSS and XSL?

Cascading Style Sheets

CSS Meet-Up
Tucson AZ, USA 5 February 2013

Status

No plans for a new language yet
- (Disappointed? Or reassured? ☺)
People ask me this sometimes
There are problems in XSL and CSS
At some point a new language may be cheaper
- or two
  - or even three

Problems in CSS can often be fixed by adding new features, but old features cannot be removed or changed, so the language is getting more and more complicated to learn, to implement and to define.

(The currently proposed @supports feature may help a bit by allowing certain things to change, but at the cost of even more complexity.)

If, as it currently looks, we will have to add quite a few new features still, then using another language (i.e., another MIME type) at some point becomes necessary.

Some CSS problems: inconsistencies

Just as examples:

`:` (`:first-line, :before`)	vs	`::` (`::marker, ::footnote-call`)
`nth-child(2n+1)` no spaces required	vs	`calc(em + 2p)` spaces required
`flex-direction: row` needs `display: flex`	vs	`columns: 2` any block container
`@supports A or B`	vs	`@media A, B`

The working group is aware of these inconsistencies, so they are not oversights. In fact, they have all been extensively and repeatedly discussed. Some may even still be reversed. The point is just that inconsistencies are more and more difficult to avoid and there will be more as more features are added that weren't originally expected for CSS.

Some CSS problems: redundancies

Relative positioning and geometric translations
SASS, LESS, etc. and variables
Absolute positioning and grid templates
'box-width' and 'calc()'
'page-break-after' and 'break-after'

Difficulties with development

Not enough resources to develop XSL-FO version 2
Bugs that last too long, causing loss of functionality
(absolute units, horizontal centering…)
Adding features to CSS takes too long, e.g:
math, page templates, element-based templates, spanning columns, footnotes, multiple footnote streams, line grid, aligning the last/only line, stretch font size to fill space, region-based styling, cross-references (“see page 7”), stretch-text, vertical alignment in columns…

E.g., the first demos of math in HTML were already shown in 1995, but it's very recent that browsers have started supporting it. And you can only control a few aspects of the display with CSS, because math typography isn't integrated into CSS.

Lack of tools

Why did it take so long for Compass, LESS, etc. to appear?

Evidence that they weren't needed?

Or that CSS is too complex for automation?

New language(s) for which users?

Occasional users
- Avoid indirections & nesting, automate typographic details, predefined templates…
Advanced/professional users
- Low-level controls, arbitrary layout, control over time vs quality…
Machines
- Think: “easy to parse with Perl”

How many new languages?

Replace CSS and/or XSL-FO? Complement them?
Two language with common “engine”
Alternative syntax for machines? (with standard mapping)
- XML? RDF?
- Cf. RDF (3 syntaxes), ESIS/Infoset (nsgmls, hxpipe)

Design principles

"Principle of least power" (Tim Berners-Lee)
- Allow users to use their favorite document management tool
- Room for some meta-data? (author, date, XMP, assertions…)
Extensibility (i.e., a 2nd version thet doesn't break older implementations)
Specify goals, assume that algorithms will improve
For advanced language: primitives for user-defined extensions
(still machine-readable)
Content-adaptation & privacy

The end

http://www.w3.org/Talks/2013/0205-Style-Tucson

W3C

To Lead the Web to its full potential

To Anticipate the Trends

To Increase your company value

Join W3C

http://www.w3.org/Consortium/join

or contact: Alan Bird <abird@w3.org>

Bert Bos <bert@w3.org>
GPG fingerprint: 7744 0204 52A5 14D9 147D 2A13 2D7A E420 184B 5BA4